Scholarlink

<p>In this paper, a dataset consisting of 112 rock specimens was used to develop predictive models for uniaxial compressive strength based on density, P-wave velocity, porosity, and Leeb hardness. Four machine learning algorithms, including M5P, multi-layer perceptron, Instance-Based K-Nearest Neighbors, and KStar, were evaluated and compared. Model performance was assessed using the correlation coefficient, root mean square error, and mean absolute error. The results showed that all models successfully captured the relationship between the input parameters and uniaxial compressive strength, although their predictive capabilities differed considerably. Among the investigated algorithms, KStar produced the most accurate predictions, yielding a correlation coefficient of 0.93, a root mean square error of 19.37, and a mean absolute error of 14.53. The superior performance of KStar indicated its effectiveness in modeling the complex nonlinear interactions between rock properties and strength. The findings demonstrated that machine learning techniques can provide reliable estimates of uniaxial compressive strength using readily measurable non-destructive testing parameters.</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

Comparative Evaluation of Intelligent Machine Learning Models for Prediction of Uniaxial Compressive Strength of Rocks

  • 期刊论文

  • 发表日期：

    2026年7月18日

查看全文
  • Sasan Ghorbani

摘要

In this paper, a dataset consisting of 112 rock specimens was used to develop predictive models for uniaxial compressive strength based on density, P-wave velocity, porosity, and Leeb hardness. Four machine learning algorithms, including M5P, multi-layer perceptron, Instance-Based K-Nearest Neighbors, and KStar, were evaluated and compared. Model performance was assessed using the correlation coefficient, root mean square error, and mean absolute error. The results showed that all models successfully captured the relationship between the input parameters and uniaxial compressive strength, although their predictive capabilities differed considerably. Among the investigated algorithms, KStar produced the most accurate predictions, yielding a correlation coefficient of 0.93, a root mean square error of 19.37, and a mean absolute error of 14.53. The superior performance of KStar indicated its effectiveness in modeling the complex nonlinear interactions between rock properties and strength. The findings demonstrated that machine learning techniques can provide reliable estimates of uniaxial compressive strength using readily measurable non-destructive testing parameters.