Scholarlink

<p>This study investigates the influence of mineralogical and textural characteristics on the mechanical properties of sandstones using machine learning algorithms. Twenty sandstone samples from four geological formations in Iran were analyzed for petrographic properties (mineralogy coefficient, MC; texture coefficient, TC) and mechanical parameters (Schmidt hardness, Hs; point load index, PLI; Brazilian tensile strength, BTS; uniaxial compressive strength, UCS; elasticity modulus, E). Thin-section, X-ray diffraction, scanning electron microscopy analyses, and standardized mechanical tests (ISRM/ASTM) were conducted, followed by predictive modeling using Linear Regression (LR), Support Vector Machine (SVM), and Gradient Boosting (GB) in Orange 3.39.0 software. Results demonstrated strong correlations between TC and mechanical properties (R² = 0.84–0.90), outperforming MC (R² = 0.47–0.57). Gradient Boosting achieved the highest predictive accuracy (RMSE = 0.13–5.31; R² ≤ 0.96), while SVM showed instability for UCS predictions (R² = −0.18). Sensitivity analysis revealed TC as the dominant predictor, with PLI-TC ranked highest across evaluation methods. Discussions highlighted texture (grain interlocking, orientation) as more influential than mineralogy alone, though combined MC-TC models improved UCS prediction by 12–15%. The study concludes that TC-based machine learning algorithms, particularly GB, offer robust tools for estimating sandstone mechanical properties from petrographic data, benefiting engineering applications where direct testing is impractical.</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

A comparative study of the mineralogy and texture effects on the sandstone mechanical properties using some machine learning algorithms

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月30日

查看全文
  • Davood Fereidooni,
  • Matloob Hejazifar

摘要

This study investigates the influence of mineralogical and textural characteristics on the mechanical properties of sandstones using machine learning algorithms. Twenty sandstone samples from four geological formations in Iran were analyzed for petrographic properties (mineralogy coefficient, MC; texture coefficient, TC) and mechanical parameters (Schmidt hardness, Hs; point load index, PLI; Brazilian tensile strength, BTS; uniaxial compressive strength, UCS; elasticity modulus, E). Thin-section, X-ray diffraction, scanning electron microscopy analyses, and standardized mechanical tests (ISRM/ASTM) were conducted, followed by predictive modeling using Linear Regression (LR), Support Vector Machine (SVM), and Gradient Boosting (GB) in Orange 3.39.0 software. Results demonstrated strong correlations between TC and mechanical properties (R² = 0.84–0.90), outperforming MC (R² = 0.47–0.57). Gradient Boosting achieved the highest predictive accuracy (RMSE = 0.13–5.31; R² ≤ 0.96), while SVM showed instability for UCS predictions (R² = −0.18). Sensitivity analysis revealed TC as the dominant predictor, with PLI-TC ranked highest across evaluation methods. Discussions highlighted texture (grain interlocking, orientation) as more influential than mineralogy alone, though combined MC-TC models improved UCS prediction by 12–15%. The study concludes that TC-based machine learning algorithms, particularly GB, offer robust tools for estimating sandstone mechanical properties from petrographic data, benefiting engineering applications where direct testing is impractical.