Scholarlink

<p>Unpaved low-volume roads are widely employed to meet both temporary and permanent transportation needs. To reinforce the subgrade soil, geosynthetic materials are commonly used, and one crucial design criterion in this context is the California Bearing Ratio (CBR), which assesses subgrade soil strength. This study focuses on the development of four hybrid intelligent models by combining the artificial neural network (ANN) with Harris hawk optimization (ANN-HHO), Equilibrium optimization (ANN-EO), dragonfly optimization (ANN-DFO) and grey wolf optimization (ANN-GWO) to forecast the CBR of geosynthetic-reinforced subgrade soil (GRSS). The pertinent historical dataset was utilized to train and test the hybrid models. Experimental results showed that the ANN-HHO model has obtained the highest predictive ability in comparison to the other models. In the validation stage and in the uncertainty analysis, the ANN-HHO model showed the best prediction performance (R<sup>2</sup> = 0.978 and RMSE = 1.484). The predictive veracity and generalization capability of the ANN-HHO model was also corroborated using external and independent validation and monotonicity analysis, respectively. Also, the ANN-HHO model for estimating the CBR of GRSS has been implemented into a user-friendly Graphical User Interface (GUI). Finally, the impact of the input parameters on the dependent variable (CBR) was evaluated via sensitivity analysis.</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

AI empowered pavement engineering: forecasting strength of geosynthetic-reinforced soils in unpaved roads

  • 期刊论文

  • 发表日期：

    2026年6月1日

查看全文
  • Muhammad Nouman Amjad Raja,
  • Tarek Abdoun,
  • Waleed El-Sekelly

摘要

Unpaved low-volume roads are widely employed to meet both temporary and permanent transportation needs. To reinforce the subgrade soil, geosynthetic materials are commonly used, and one crucial design criterion in this context is the California Bearing Ratio (CBR), which assesses subgrade soil strength. This study focuses on the development of four hybrid intelligent models by combining the artificial neural network (ANN) with Harris hawk optimization (ANN-HHO), Equilibrium optimization (ANN-EO), dragonfly optimization (ANN-DFO) and grey wolf optimization (ANN-GWO) to forecast the CBR of geosynthetic-reinforced subgrade soil (GRSS). The pertinent historical dataset was utilized to train and test the hybrid models. Experimental results showed that the ANN-HHO model has obtained the highest predictive ability in comparison to the other models. In the validation stage and in the uncertainty analysis, the ANN-HHO model showed the best prediction performance (R2 = 0.978 and RMSE = 1.484). The predictive veracity and generalization capability of the ANN-HHO model was also corroborated using external and independent validation and monotonicity analysis, respectively. Also, the ANN-HHO model for estimating the CBR of GRSS has been implemented into a user-friendly Graphical User Interface (GUI). Finally, the impact of the input parameters on the dependent variable (CBR) was evaluated via sensitivity analysis.