Scholarlink

<p>Dissolution channels significantly affect the mechanical behavior of karst slopes. Triaxial compression tests, 3D X-ray CT scanning, and machine learning are employed to investigate the failure characteristics and crack evolution in multi-hole limestone. A random forest model was developed to predict the crack fractal dimension, enabling a quantitative assessment of post-failure crack complexity. Results indicate that holed limestone develops through-going shear cracks and intersecting tensile cracks, with crack propagation increasing in quantity, length, and uniformity as hole number rises. The proportion of short cracks decreased from 69.50% to 50.70%, while ultra-long cracks increased from 35 to 93, and maximum crack area expanded by 68.37%. Peak and residual strength declined quadratically with increasing CT volume fracture (<i>V</i>ct); when <i>V</i>ct rose from 5.5% to 10.8%, peak and residual strength dropped by 22.32% and 46.07%, respectively. A strong positive correlation observed between <i>V</i>ct and 3D fractal dimension (<i>D</i>s). The proposed random forest model achieved MSE, RMSE, and MAE values below 0.02, with <i>R</i>² &gt;0.9, demonstrating high predictive accuracy. These findings provide a robust foundation for rock stability assessment in karst regions, offering insights into failure mechanisms and disaster prevention strategies.</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

Quantifying crack and fractal features in multi-hole limestone combined with x-ray CT and machine learning

  • 期刊论文

  • 发表日期：

    2025年10月22日

查看全文
  • Zuliang Zhong,
  • Kaixin Zhu,
  • Nanyun Wang,
  • Laiyang Li,
  • Zezhou Li

摘要

Dissolution channels significantly affect the mechanical behavior of karst slopes. Triaxial compression tests, 3D X-ray CT scanning, and machine learning are employed to investigate the failure characteristics and crack evolution in multi-hole limestone. A random forest model was developed to predict the crack fractal dimension, enabling a quantitative assessment of post-failure crack complexity. Results indicate that holed limestone develops through-going shear cracks and intersecting tensile cracks, with crack propagation increasing in quantity, length, and uniformity as hole number rises. The proportion of short cracks decreased from 69.50% to 50.70%, while ultra-long cracks increased from 35 to 93, and maximum crack area expanded by 68.37%. Peak and residual strength declined quadratically with increasing CT volume fracture (Vct); when Vct rose from 5.5% to 10.8%, peak and residual strength dropped by 22.32% and 46.07%, respectively. A strong positive correlation observed between Vct and 3D fractal dimension (Ds). The proposed random forest model achieved MSE, RMSE, and MAE values below 0.02, with R² >0.9, demonstrating high predictive accuracy. These findings provide a robust foundation for rock stability assessment in karst regions, offering insights into failure mechanisms and disaster prevention strategies.