Click the blue "muyitong" above
Take the lead in obtaining massive material data free inquiry service!
Prediction of density and elastic modulus of silica based glass based on machine learning
Predictingdensities and elastic moduli of SiO2-based glasses by machine learning
Yong-Jie Hu, Ge Zhao, Mingfei Zhang, Bin Bin, Tyler Del Rose, QianZhao, Qun Zu, Yang Chen, Xuekun Sun, Maarten de Jong and Liang Qi
npjComputational Materials (2020) 6:25
https://doi.org/10.1038/s41524-020-0291-z
Abstract
The SiO2 based glass with high elastic modulus and light weight obtained by chemical design has received great attention. However, because the elastic modulus is a complex function of the bond between atoms and its order at different length scales, it is difficult to find a general expression to predict the elastic modulus of glass before synthesis. This study shows that the density and elastic modulus of SiO2 based glass can be predicted by machine learning method in the complex composition combination with various oxide additives. The machine learning method is based on the training set obtained by high-throughput molecular dynamics simulation. The training set unifies the empirical statistical model based on the basic physics of atomic bonding and the statistical learning / prediction model based on the minimum absolute shrinkage and selection operator (gbm-lasso) of the iterative decision tree. Machine learning prediction has been comprehensively compared and verified by a large number of simulation and experimental data. Through the training of binary and triplet glass sample data set, the model shows excellent ability to predict the density and elasticity of SiO2 based glass beyond the training set. As an example of its potential application, gbm-lasso model is used to quickly and cheaply screen various components (~ 105) of multi-component glass system, so as to build a component performance database to comprehensively evaluate the density and elasticity of glass.
Accompanying article
Figure 1 performance of machine learning model in glass density prediction
Fig. 2 prediction performance of machine learning model for glass components outside the training set
Figure 3 the expansibility evaluation of gbm-lasso model for prediction of new oxide system
Figure 4 prediction of glass properties by gbm-lasso model based on experimental data
Figure 5 distribution of density and Young's modulus in na2o-cao-al2o3-y2o3-sio2 system
Matai products take material genetic engineering technology as the framework, integrate material Informatics, data mining, machine learning and artificial intelligence technology, accelerate the research and development and screening of new materials, assist high-tech enterprises and scientific research institutions, and carry out material design, screening and discovery with low threshold, simple operation and high efficiency using material big data technology.
Matai uses idatacenter system to build a systematic, structured, complete and standardized material database for customers, and uses idatainsight system to build a solution for customers to study material machine learning. IData insight adopts appropriate model for training and optimization to realize the visual exploration and discovery of the law of "material composition process organization performance service".
Www.mat.ai
Leader in materials digital solutions
More details (click)