Scholarlink

<p>Lithium titanate (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\:{\varvec{L}\varvec{i}}_{2}\varvec{T}\varvec{i}{\varvec{O}}_{3}\)</EquationSource> </InlineEquation>, LT) has emerged as a promising material owing to several favorable characteristics, including superior performance, structural stability, affordability, and low toxicity. As a result, it is an excellent candidate for use in energy storage and conversion applications. The mechanical properties, including elastic anisotropy, shear modulus, bulk modulus, and Poisson’s ratio, are critical for assessing the material’s structural integrity and durability under operating conditions, particularly in demanding applications such as lithium-ion batteries and fusion reactors. Here, we benchmark three supervised learning models (linear regression, random forest, and XGBoost) to predict these properties from simple structural descriptors, including unit-cell volume and the number of sites. Despite the limited dataset size, the models achieve strong predictive accuracy, highlighting the potential of machine learning to accelerate mechanical-property screening and support materials optimization.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting the mechanical properties of \(\:{Li}_{2}Ti{O}_{3}\) using machine learning process: linear regression, random forests and XGBoost models

  • 期刊论文

  • 发表日期：

    2026年2月17日

查看全文
  • Housna Dari,
  • Achraf Chahbi,
  • Fatima Zahra Krimech,
  • Ahmed Hader,
  • Ilias Tarras,
  • Rachida Moultif,
  • Mohammed Tanasehte,
  • Yahia Boughaleb

摘要

Lithium titanate ( \(\:{\varvec{L}\varvec{i}}_{2}\varvec{T}\varvec{i}{\varvec{O}}_{3}\) , LT) has emerged as a promising material owing to several favorable characteristics, including superior performance, structural stability, affordability, and low toxicity. As a result, it is an excellent candidate for use in energy storage and conversion applications. The mechanical properties, including elastic anisotropy, shear modulus, bulk modulus, and Poisson’s ratio, are critical for assessing the material’s structural integrity and durability under operating conditions, particularly in demanding applications such as lithium-ion batteries and fusion reactors. Here, we benchmark three supervised learning models (linear regression, random forest, and XGBoost) to predict these properties from simple structural descriptors, including unit-cell volume and the number of sites. Despite the limited dataset size, the models achieve strong predictive accuracy, highlighting the potential of machine learning to accelerate mechanical-property screening and support materials optimization.