Scholarlink

<p>This study investigated the efficiency of sugarcane bagasse ash (SCBA) and nano-silica (NS) in improving the California Bearing Ratio (CBR) of sandy subgrade soil, with predictions modelled using machine learning (ML) algorithms. The objective was to evaluate the individual and combined effects of these sustainable stabilisers on CBR under varying curing periods, and to develop accurate, interpretable ML models for CBR prediction that can support sustainable pavement design. A total of 480 laboratory-prepared soil specimens were tested under unsoaked and soaked conditions, incorporating varying SCBA contents (5–15%), NS contents (1–2%), and curing periods (0–28 days). Atterberg limits, optimum moisture content, and maximum dry density (MDD) were evaluated. Six ML algorithms, including Random Forest, Support Vector Machine, Light Gradient Boosting Machine (Light GBM), Extreme Gradient Boosting, Decision Tree, and K-Nearest Neighbors, were trained on the dataset using a 70:30 train-test split and stratified 10-fold cross-validation. Model performance was evaluated using Mean Absolute Error, Mean Squared Error, Root Mean Squared Error, Root Mean Squared Logarithmic Error, and Coefficient of Determination metrics, with SHapley Additive exPlanations (SHAP) applied for feature importance ranking. Results indicated significant CBR improvements, with maximum unsoaked and soaked values of 27.82% and 20.22%, respectively, achieved at 15% SCBA + 2% NS after 28 days. Light GBM outperformed other models in unsoaked and soaked conditions. SHAP analysis identified the curing period as the dominant feature, followed by MDD, plastic limit, and NS. In conclusion, SCBA and NS offer sustainables stabilisation for sandy subgrades, with ensemble ML models providing reliable, explainable predictions.</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

Modelling Unsoaked and Soaked California Bearing Ratio of Nano-Stabilised Sandy Subgrade Soil Using Machine Learning Algorithms

  • 期刊论文

  • 发表日期：

    2025年10月17日

查看全文
  • Neero Gumsar Sorum,
  • Pradeep Pradhan,
  • Joana Rikchi D. Sangma

摘要

This study investigated the efficiency of sugarcane bagasse ash (SCBA) and nano-silica (NS) in improving the California Bearing Ratio (CBR) of sandy subgrade soil, with predictions modelled using machine learning (ML) algorithms. The objective was to evaluate the individual and combined effects of these sustainable stabilisers on CBR under varying curing periods, and to develop accurate, interpretable ML models for CBR prediction that can support sustainable pavement design. A total of 480 laboratory-prepared soil specimens were tested under unsoaked and soaked conditions, incorporating varying SCBA contents (5–15%), NS contents (1–2%), and curing periods (0–28 days). Atterberg limits, optimum moisture content, and maximum dry density (MDD) were evaluated. Six ML algorithms, including Random Forest, Support Vector Machine, Light Gradient Boosting Machine (Light GBM), Extreme Gradient Boosting, Decision Tree, and K-Nearest Neighbors, were trained on the dataset using a 70:30 train-test split and stratified 10-fold cross-validation. Model performance was evaluated using Mean Absolute Error, Mean Squared Error, Root Mean Squared Error, Root Mean Squared Logarithmic Error, and Coefficient of Determination metrics, with SHapley Additive exPlanations (SHAP) applied for feature importance ranking. Results indicated significant CBR improvements, with maximum unsoaked and soaked values of 27.82% and 20.22%, respectively, achieved at 15% SCBA + 2% NS after 28 days. Light GBM outperformed other models in unsoaked and soaked conditions. SHAP analysis identified the curing period as the dominant feature, followed by MDD, plastic limit, and NS. In conclusion, SCBA and NS offer sustainables stabilisation for sandy subgrades, with ensemble ML models providing reliable, explainable predictions.