Machine learning prediction of glass-forming ability in bulk metallic glasses

Jie Xiong; San Qiang Shi; Tong Yi Zhang

doi:10.1016/j.commatsci.2021.110362

Machine learning prediction of glass-forming ability in bulk metallic glasses

Jie Xiong, San Qiang Shi, Tong Yi Zhang

Research output: Journal article publication › Journal article › Academic research › peer-review

36 Citations (Scopus)

Abstract

The critical casting diameter (D_max) quantitatively represents glass-forming ability (GFA) of bulk metallic glasses (BMGs). The present work constructed a dataset of two subsets, L-GFA subset of 376 BMGs with 1 mm ≤D_max < 5 mm and G-GFA subset of 319 BMGs with D_max ≥ 5 mm. The sequential backward selector and exhaustive feature selector are introduced to select key features. The trained XGBoost classifier with four selected features is able to successfully classify the L-GFA and G-GFA BMGs. Furthermore, the trained XGBoost regression model with another four selected features predicts the D_max of G-GFA samples with a cross-validated correlation coefficient of 0.8012. The correlation between features and D_max will provide the guidance in the design and discovery of novel BMGs.

Original language	English
Article number	110362
Journal	Computational Materials Science
Volume	192
DOIs	https://doi.org/10.1016/j.commatsci.2021.110362
Publication status	Published - May 2021

Keywords

Bulk metallic glasses
Glass-forming ability
Machine learning
XGBoost

ASJC Scopus subject areas

General Computer Science
General Chemistry
General Materials Science
Mechanics of Materials
General Physics and Astronomy
Computational Mathematics

More information

10.1016/j.commatsci.2021.110362

http://hdl.handle.net/10397/94236

Cite this

@article{14d2d42f16134c43a45c464af10c3c26,

title = "Machine learning prediction of glass-forming ability in bulk metallic glasses",

abstract = "The critical casting diameter (Dmax) quantitatively represents glass-forming ability (GFA) of bulk metallic glasses (BMGs). The present work constructed a dataset of two subsets, L-GFA subset of 376 BMGs with 1 mm ≤Dmax < 5 mm and G-GFA subset of 319 BMGs with Dmax ≥ 5 mm. The sequential backward selector and exhaustive feature selector are introduced to select key features. The trained XGBoost classifier with four selected features is able to successfully classify the L-GFA and G-GFA BMGs. Furthermore, the trained XGBoost regression model with another four selected features predicts the Dmax of G-GFA samples with a cross-validated correlation coefficient of 0.8012. The correlation between features and Dmax will provide the guidance in the design and discovery of novel BMGs.",

keywords = "Bulk metallic glasses, Glass-forming ability, Machine learning, XGBoost",

author = "Jie Xiong and Shi, {San Qiang} and Zhang, {Tong Yi}",

note = "Funding Information: This work was supported by the National Key Research and Development Program of China (No. 2018YFB0704400 ), the Hong Kong Polytechnic University (internal grant Nos. 1-ZE8R and G-YBDH ), and the 111 Project (grant No. D16002) from the State Administration of Foreign Experts Affairs , PRC. Publisher Copyright: {\textcopyright} 2021 Elsevier B.V. Copyright: Copyright 2021 Elsevier B.V., All rights reserved.",

year = "2021",

month = may,

doi = "10.1016/j.commatsci.2021.110362",

language = "English",

volume = "192",

journal = "Computational Materials Science",

issn = "0927-0256",

publisher = "Elsevier B.V.",

}

TY - JOUR

T1 - Machine learning prediction of glass-forming ability in bulk metallic glasses

AU - Xiong, Jie

AU - Shi, San Qiang

AU - Zhang, Tong Yi

N1 - Funding Information: This work was supported by the National Key Research and Development Program of China (No. 2018YFB0704400 ), the Hong Kong Polytechnic University (internal grant Nos. 1-ZE8R and G-YBDH ), and the 111 Project (grant No. D16002) from the State Administration of Foreign Experts Affairs , PRC. Publisher Copyright: © 2021 Elsevier B.V. Copyright: Copyright 2021 Elsevier B.V., All rights reserved.

PY - 2021/5

Y1 - 2021/5

N2 - The critical casting diameter (Dmax) quantitatively represents glass-forming ability (GFA) of bulk metallic glasses (BMGs). The present work constructed a dataset of two subsets, L-GFA subset of 376 BMGs with 1 mm ≤Dmax < 5 mm and G-GFA subset of 319 BMGs with Dmax ≥ 5 mm. The sequential backward selector and exhaustive feature selector are introduced to select key features. The trained XGBoost classifier with four selected features is able to successfully classify the L-GFA and G-GFA BMGs. Furthermore, the trained XGBoost regression model with another four selected features predicts the Dmax of G-GFA samples with a cross-validated correlation coefficient of 0.8012. The correlation between features and Dmax will provide the guidance in the design and discovery of novel BMGs.

AB - The critical casting diameter (Dmax) quantitatively represents glass-forming ability (GFA) of bulk metallic glasses (BMGs). The present work constructed a dataset of two subsets, L-GFA subset of 376 BMGs with 1 mm ≤Dmax < 5 mm and G-GFA subset of 319 BMGs with Dmax ≥ 5 mm. The sequential backward selector and exhaustive feature selector are introduced to select key features. The trained XGBoost classifier with four selected features is able to successfully classify the L-GFA and G-GFA BMGs. Furthermore, the trained XGBoost regression model with another four selected features predicts the Dmax of G-GFA samples with a cross-validated correlation coefficient of 0.8012. The correlation between features and Dmax will provide the guidance in the design and discovery of novel BMGs.

KW - Bulk metallic glasses

KW - Glass-forming ability

KW - Machine learning

KW - XGBoost

UR - http://www.scopus.com/inward/record.url?scp=85101407251&partnerID=8YFLogxK

U2 - 10.1016/j.commatsci.2021.110362

DO - 10.1016/j.commatsci.2021.110362

M3 - Journal article

AN - SCOPUS:85101407251

SN - 0927-0256

VL - 192

JO - Computational Materials Science

JF - Computational Materials Science

M1 - 110362

ER -

Machine learning prediction of glass-forming ability in bulk metallic glasses

Abstract

Keywords

ASJC Scopus subject areas

More information

Other files and links

Fingerprint

Cite this