Scholarlink

<p>Accurate assessment of rock brittleness is essential for excavation design, drilling optimization, and stability evaluation. This study proposes a reliability-based framework for predicting the punch penetration test brittleness index (BIm) from uniaxial compressive strength (UCS), Brazilian tensile strength (BTS), and density. A sparse Bayesian surrogate model—Relevance Vector Regression enhanced by Artificial Bee Colony (RVR-ABC)—is developed to provide both high prediction accuracy and calibrated uncertainty estimates. The surrogate is embedded within a hybrid Genetic Algorithm-assisted First-Order Reliability Method (GA–FORM) and Monte Carlo Simulation (MCS) to evaluate the probability of failure (PF) for critical BIm thresholds. Laboratory testing on 48 rock specimens provided the input–output dataset. The RVR-ABC achieved <i>R</i><sup>2</sup> = 0.94 and RMSE = 1.97 (BIm units) on an 80/20 test split, demonstrating strong generalization in a small-sample setting. Reliability analysis at a threshold BIm = 35 produced PF = 17.8% (MCS) and PF = 17.6% (GA–FORM), with UCS identified as the most influential parameter via FORM sensitivity measures. The proposed approach advances brittleness assessment by enabling efficient, uncertainty-aware reliability analysis that is adaptable to alternative brittleness definitions and larger, more diverse datasets, offering a practical tool for geomechanical design and risk evaluation.</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

Reliability-Based Assessment of Rock Brittleness Using Hybrid Soft Computing and Probabilistic Approaches

  • 期刊论文

  • 发表日期：

    2025年12月15日

查看全文
  • Hadi Fattahi,
  • Fatemeh Jiriyayi,
  • Danial Jahed Armaghani

摘要

Accurate assessment of rock brittleness is essential for excavation design, drilling optimization, and stability evaluation. This study proposes a reliability-based framework for predicting the punch penetration test brittleness index (BIm) from uniaxial compressive strength (UCS), Brazilian tensile strength (BTS), and density. A sparse Bayesian surrogate model—Relevance Vector Regression enhanced by Artificial Bee Colony (RVR-ABC)—is developed to provide both high prediction accuracy and calibrated uncertainty estimates. The surrogate is embedded within a hybrid Genetic Algorithm-assisted First-Order Reliability Method (GA–FORM) and Monte Carlo Simulation (MCS) to evaluate the probability of failure (PF) for critical BIm thresholds. Laboratory testing on 48 rock specimens provided the input–output dataset. The RVR-ABC achieved R2 = 0.94 and RMSE = 1.97 (BIm units) on an 80/20 test split, demonstrating strong generalization in a small-sample setting. Reliability analysis at a threshold BIm = 35 produced PF = 17.8% (MCS) and PF = 17.6% (GA–FORM), with UCS identified as the most influential parameter via FORM sensitivity measures. The proposed approach advances brittleness assessment by enabling efficient, uncertainty-aware reliability analysis that is adaptable to alternative brittleness definitions and larger, more diverse datasets, offering a practical tool for geomechanical design and risk evaluation.