Scholarlink

<p>When the pavement layers get inundated, the rebounding reflux get affected and gradually the layer stiffens or permanently deforms under dynamic loads. To comprehend the deformation properties of the layer, it becomes essential to record the resilient response of the layer as a result of moisture interventions. The reactiveness of compacted soil mass can be valued based on the index termed Resilient Modulus (M<sub>R</sub>). These responses especially under unsaturated states are highly unpredictable and thus if a soil specific variability data of M<sub>R</sub> is available, it can ease the design process. The main objective is to predict the resilient response, M<sub>R</sub>, of laterite soils at different moisture conditions, using advanced computing tools like Artificial Neural Network (ANN), Gaussian Process Regression (GPR) and Support Vector machine (SVM). In order to develop dataset for computation, resilient modulus models with soil suction (Ψ) parameter and Environmental modification factor (F<sub>U</sub>) were utilized. It was observed that the soft computing techniques ANN, GPR and SVM could fairly predict the M<sub>R</sub> of laterite soils at variable moisture and stress properties with R<sup>2</sup> of 0.98, 0.94 and 0.89 respectively, when confining pressure (kPa), deviator stress (kPa), S-S<sub>opt</sub> (%), percent of fines and plasticity index (%) were provided as input to the system. It was observed that the neural networking tool could predict comparatively better than GPR and SVR models.</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

Environmental factor modified resilient modulus prediction of tropical fine grained laterite soils using artificial neural network, Gaussian process and support vector approaches

  • 期刊论文

  • 发表日期：

    2025年12月9日

查看全文
  • Soorya Ann Koshy,
  • U. Salini

摘要

When the pavement layers get inundated, the rebounding reflux get affected and gradually the layer stiffens or permanently deforms under dynamic loads. To comprehend the deformation properties of the layer, it becomes essential to record the resilient response of the layer as a result of moisture interventions. The reactiveness of compacted soil mass can be valued based on the index termed Resilient Modulus (MR). These responses especially under unsaturated states are highly unpredictable and thus if a soil specific variability data of MR is available, it can ease the design process. The main objective is to predict the resilient response, MR, of laterite soils at different moisture conditions, using advanced computing tools like Artificial Neural Network (ANN), Gaussian Process Regression (GPR) and Support Vector machine (SVM). In order to develop dataset for computation, resilient modulus models with soil suction (Ψ) parameter and Environmental modification factor (FU) were utilized. It was observed that the soft computing techniques ANN, GPR and SVM could fairly predict the MR of laterite soils at variable moisture and stress properties with R2 of 0.98, 0.94 and 0.89 respectively, when confining pressure (kPa), deviator stress (kPa), S-Sopt (%), percent of fines and plasticity index (%) were provided as input to the system. It was observed that the neural networking tool could predict comparatively better than GPR and SVR models.