Scholarlink

<p>To investigate the factors controlling loess shear strength and accurately predict its variation, unconsolidated–undrained (UU) triaxial tests were conducted under six water content levels and three confining pressures. Based on the experimental results, a loess shear strength database was established using the finite element method (FEM), and an artificial neural network (ANN) model was developed with six input parameters: cohesion, internal friction angle, confining pressure, density, elastic modulus, and Poisson’s ratio. The results show that water content and confining pressure are the dominant factors governing loess shear strength. At a confining pressure of 100&#xa0;kPa, increasing the water content from 15.5% to 19.5% leads to a 60% reduction in failure deviatoric stress, whereas at a water content of 19.5%, increasing the confining pressure from 100&#xa0;kPa to 300&#xa0;kPa results in a 400% increase. FEM simulation results agree well with triaxial test data, confirming the reliability of the constructed dataset. The ANN model demonstrates high predictive accuracy and strong genferalization capability, with predicted deviatoric stresses closely matching experimental values (<i>R</i>² = 0.9897, <i>RMSE</i> = 0.0915 for the entire dataset). Sensitivity analysis indicates that cohesion, internal friction angle, density, and confining pressure play major roles in controlling shear strength, while elastic modulus and Poisson’s ratio have negligible effects.</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

Prediction of loess shear strength using triaxial test and artificial neural network

  • 期刊论文

  • 发表日期：

    2026年4月28日

查看全文
  • Zhiyi Huang,
  • Jianming Ling,
  • Shiye Chen,
  • Qinglong You,
  • Mi Xiong,
  • Wenxu Huang

摘要

To investigate the factors controlling loess shear strength and accurately predict its variation, unconsolidated–undrained (UU) triaxial tests were conducted under six water content levels and three confining pressures. Based on the experimental results, a loess shear strength database was established using the finite element method (FEM), and an artificial neural network (ANN) model was developed with six input parameters: cohesion, internal friction angle, confining pressure, density, elastic modulus, and Poisson’s ratio. The results show that water content and confining pressure are the dominant factors governing loess shear strength. At a confining pressure of 100 kPa, increasing the water content from 15.5% to 19.5% leads to a 60% reduction in failure deviatoric stress, whereas at a water content of 19.5%, increasing the confining pressure from 100 kPa to 300 kPa results in a 400% increase. FEM simulation results agree well with triaxial test data, confirming the reliability of the constructed dataset. The ANN model demonstrates high predictive accuracy and strong genferalization capability, with predicted deviatoric stresses closely matching experimental values (R² = 0.9897, RMSE = 0.0915 for the entire dataset). Sensitivity analysis indicates that cohesion, internal friction angle, density, and confining pressure play major roles in controlling shear strength, while elastic modulus and Poisson’s ratio have negligible effects.