Scholarlink

<p>Young's modulus (E) is a crucial parameter for predicting a material's ability to withstand pressure and is essential in designing rock engineering projects. E has wide applications in mining, geotechnical engineering, and other fields. While E can be measured directly through laboratory tests, this requires high-quality core samples and expensive modern equipment. Therefore, an indirect method for estimating E is an attractive alternative. In this study, four novel data-driven approaches—ridge regression (RR), Lasso regression (LR), artificial neural network (ANN), and gradient boosting regressor (GBR)—were developed to predict E. The dataset of E was divided into 70% for training and 30% for testing for each model. To improve the performance of each model, an iterative fivefold cross-validation method was used. Results showed that the GBR regression model outperformed the other models, achieving higher accuracy with a correlation coefficient (R<sup>2</sup>) of 0.995 on the training set and 0.992 on the testing set, mean absolute errors (MAE) of 0.0162 and 0.0147, respectively, and root mean square errors (RMSE) of 0.02 and 0.0173. The model also scored highly on the <i>a</i>20-index, with 0.96 at training and 0.98 at testing. Using these four regression models, this study provides alternative methods to predict E accurately and efficiently.</p><p><b>Highlights</b><UnorderedList Mark="Bullet"> <ItemContent> <p>This study presents the first comparative implementation of four data-driven models—Ridge Regression, Lasso Regression, Artificial Neural Networks, and Gradient Boosting Regression—for predicting rock elastic modulus, establishing a new methodological framework for geomechanical parameter estimation.</p> </ItemContent> <ItemContent> <p>The proposed Gradient Boosting Regressor demonstrated exceptional predictive accuracy with a test R<sup>2</sup> of 0.990, significantly outperforming other models in capturing complex nonlinear relationships between rock properties and elastic modulus.</p> </ItemContent> <ItemContent> <p>Through systematic sensitivity analysis, Uniaxial Compressive Strength and Brazilian Tensile Strength were identified as the most influential parameters, providing crucial insights for optimizing field data collection in rock engineering applications.</p> </ItemContent> </UnorderedList></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 Data-Driven Method for Predicting Rocks' Young’s Modulus: Case Study

  • 期刊论文

  • 发表日期：

    2026年3月13日

查看全文
  • Wei Xin,
  • Jun Li,
  • Xigui Zheng,
  • Xiaowei Qin

摘要

Young's modulus (E) is a crucial parameter for predicting a material's ability to withstand pressure and is essential in designing rock engineering projects. E has wide applications in mining, geotechnical engineering, and other fields. While E can be measured directly through laboratory tests, this requires high-quality core samples and expensive modern equipment. Therefore, an indirect method for estimating E is an attractive alternative. In this study, four novel data-driven approaches—ridge regression (RR), Lasso regression (LR), artificial neural network (ANN), and gradient boosting regressor (GBR)—were developed to predict E. The dataset of E was divided into 70% for training and 30% for testing for each model. To improve the performance of each model, an iterative fivefold cross-validation method was used. Results showed that the GBR regression model outperformed the other models, achieving higher accuracy with a correlation coefficient (R2) of 0.995 on the training set and 0.992 on the testing set, mean absolute errors (MAE) of 0.0162 and 0.0147, respectively, and root mean square errors (RMSE) of 0.02 and 0.0173. The model also scored highly on the a20-index, with 0.96 at training and 0.98 at testing. Using these four regression models, this study provides alternative methods to predict E accurately and efficiently.

Highlights

This study presents the first comparative implementation of four data-driven models—Ridge Regression, Lasso Regression, Artificial Neural Networks, and Gradient Boosting Regression—for predicting rock elastic modulus, establishing a new methodological framework for geomechanical parameter estimation.

The proposed Gradient Boosting Regressor demonstrated exceptional predictive accuracy with a test R2 of 0.990, significantly outperforming other models in capturing complex nonlinear relationships between rock properties and elastic modulus.

Through systematic sensitivity analysis, Uniaxial Compressive Strength and Brazilian Tensile Strength were identified as the most influential parameters, providing crucial insights for optimizing field data collection in rock engineering applications.