Scholarlink

<p>By clustering test sound signals, an impending failure sound (IFS) signal was uncovered, whose occurrence serves as a new sign of impending hard rock failure. These tests encompassed five classic tests (uniaxial, biaxial, true-triaxial, direct shear, Brazilian disk) across three lithologies (granite, limestone, sandstone), supplemented by the one free face and five stressed faces, true-triaxial weak dynamic disturbance and true-triaxial unloading tests, yielding 18 test sets, each with 3 repetitions. Sound features were clustered using an unsupervised learning method with the unsupervised k-means, hybrid fuzzy C-means and particle swarm optimization (HFP), and First Integer Neighbor Clustering Hierarchy (FINCH) algorithms, followed by comparative analysis. The analysis revealed that the IFS signal precedes hard rock failure: It was observed in 18/18 sets with U-k-means, 13/18 with HFP, and 11/18 with FINCH. The mean IFS lead time is 0.11 of the total loading time (95% confidence interval: 0.09–0.13). In feature space, the IFS signal exhibits a high zero-crossing rate, high spectral entropy, an elevated linear spectral centroid, broadened Mel- and Bark-band spreads, and high amplitude. Unlike trend-type signs (e.g., <i>b</i> value or fractal dimension), the new sign is event-type, which can be readily separated from other signals using these six features. Moreover, the IFS signal encodes the structural transition in hard rock failure from dispersed micro-fracture to through-going, cooperative fracture. Therefore, its occurrence is algorithm-agnostic and robust across hard rock types and stress paths. Applying this sign for hard rock failure early warning avoids the difficulties associated with measuring on-site stress and historical stress paths.</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

A robust sound-based sign of impending hard rock failure

  • 期刊论文

  • 发表日期：

    2025年11月26日

查看全文
  • Guoshao Su,
  • Xianjie Chen,
  • Zonghui Liu,
  • Peifeng Li,
  • Jianqing Jiang

摘要

By clustering test sound signals, an impending failure sound (IFS) signal was uncovered, whose occurrence serves as a new sign of impending hard rock failure. These tests encompassed five classic tests (uniaxial, biaxial, true-triaxial, direct shear, Brazilian disk) across three lithologies (granite, limestone, sandstone), supplemented by the one free face and five stressed faces, true-triaxial weak dynamic disturbance and true-triaxial unloading tests, yielding 18 test sets, each with 3 repetitions. Sound features were clustered using an unsupervised learning method with the unsupervised k-means, hybrid fuzzy C-means and particle swarm optimization (HFP), and First Integer Neighbor Clustering Hierarchy (FINCH) algorithms, followed by comparative analysis. The analysis revealed that the IFS signal precedes hard rock failure: It was observed in 18/18 sets with U-k-means, 13/18 with HFP, and 11/18 with FINCH. The mean IFS lead time is 0.11 of the total loading time (95% confidence interval: 0.09–0.13). In feature space, the IFS signal exhibits a high zero-crossing rate, high spectral entropy, an elevated linear spectral centroid, broadened Mel- and Bark-band spreads, and high amplitude. Unlike trend-type signs (e.g., b value or fractal dimension), the new sign is event-type, which can be readily separated from other signals using these six features. Moreover, the IFS signal encodes the structural transition in hard rock failure from dispersed micro-fracture to through-going, cooperative fracture. Therefore, its occurrence is algorithm-agnostic and robust across hard rock types and stress paths. Applying this sign for hard rock failure early warning avoids the difficulties associated with measuring on-site stress and historical stress paths.