Scholarlink

<p>This study evaluates machine learning models for predicting bearing capacity of geogrid-reinforced stone columns in soft clay. Using 246 samples from lab tests with 50&#xa0;mm diameter columns, three models were compared: Gaussian Process Regression (GPR), Support Vector Machine (SVM), and Genetic Algorithm-optimized SVM (GA-SVM). Correlation analysis showed bearing capacity of unreinforced clay (r = 0.79) and settlement-to-diameter ratio (r = 0.77) were most influential, while thickness and length-to-diameter ratios had minimal impact. GPR achieved superior prediction accuracy (R<sup>2</sup> = 0.9999/0.9998 for training/testing) compared to SVM (R<sup>2</sup> = 0.9972) and GA-SVM (R<sup>2</sup> = 0.9985). Genetic algorithm optimization improved standard SVM by reducing testing RMSE from 4.1606&#xa0;kPa to 3.0689&#xa0;kPa.</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

Machine Learning Approaches for Predicting Bearing Capacity of Geogrid-Reinforced Stone Columns in Soft Clay

  • 期刊论文

  • 发表日期：

    2026年4月9日

查看全文
  • Pranshu Vardhan,
  • Rakesh Kumar,
  • Suneet Kaur

摘要

This study evaluates machine learning models for predicting bearing capacity of geogrid-reinforced stone columns in soft clay. Using 246 samples from lab tests with 50 mm diameter columns, three models were compared: Gaussian Process Regression (GPR), Support Vector Machine (SVM), and Genetic Algorithm-optimized SVM (GA-SVM). Correlation analysis showed bearing capacity of unreinforced clay (r = 0.79) and settlement-to-diameter ratio (r = 0.77) were most influential, while thickness and length-to-diameter ratios had minimal impact. GPR achieved superior prediction accuracy (R2 = 0.9999/0.9998 for training/testing) compared to SVM (R2 = 0.9972) and GA-SVM (R2 = 0.9985). Genetic algorithm optimization improved standard SVM by reducing testing RMSE from 4.1606 kPa to 3.0689 kPa.