东北大学学报(自然科学版) ›› 2025, Vol. 46 ›› Issue (3): 130-137.DOI: 10.12068/j.issn.1005-3026.2025.20230259

• 资源与土木工程 • 上一篇    下一篇

改进的密度峰值聚类算法在岩体结构面优势分组中的应用

王述红, 高晨翔(), 侯钦宽   

  1. 东北大学 资源与土木工程学院,辽宁 沈阳 110819
  • 收稿日期:2023-09-04 出版日期:2025-03-15 发布日期:2025-05-29
  • 通讯作者: 高晨翔
  • 作者简介:王述红(1969―),男,江苏泰州人,东北大学教授,博士生导师.
  • 基金资助:
    中国-中东欧国家高校联合教育项目(2021111);国家自然科学基金资助项目(U1602232);辽宁省重点科技计划项目(2019JH2-10100035);中央高校基本科研业务费(N2301005)

Application of Improved Density Peak Clustering Algorithm in Dominant Grouping of Rock Discontinuities

Shu-hong WANG, Chen-xiang GAO(), Qin-kuan HOU   

  1. School of Resources & Civil Engineering,Northeastern University,Shenyang 110819,China.
  • Received:2023-09-04 Online:2025-03-15 Published:2025-05-29
  • Contact: Chen-xiang GAO

摘要:

岩体稳定性评价依赖于合理的结构面分组,但传统方法存在易受边缘点与异常点影响的弊端.为此,提出一种改进的密度峰值聚类算法用于结构面优势分组.首先,将结构面产状转换为空间坐标,并以单位法向量夹角正弦值的平方作为相似性度量.随后,基于有效性评价指标构建目标函数,并利用乌鸦算法优化截断距离以获取最佳分组结果.通过模拟数据集验证了该算法能够有效减少人为干预,避免异常点干扰,确保聚类结果更加可靠和合理.结果表明,所提方法不仅与传统方法一致性良好,还具有更高的适用性,为工程中结构面优势分组提供了可靠的参考.

关键词: 密度峰值聚类, 乌鸦算法, 有效性评价指标, 结构面, 优势分组

Abstract:

The stability evaluation of rock mass relies on reasonable rock discontinuities grouping. However, traditional methods are susceptible to boundary and outlier points. To address this issue, an improved density peak clustering algorithm was proposed for rock discontinuities grouping. Firstly, the rock discontinuities orientations were converted into spatial coordinates, and the squared sine of the angle between unit normal vectors was used as a similarity metric. Then, an objective function was constructed based on validity evaluation indices, and the cutoff distance was optimized using the crow algorithm to obtain the optimal grouping results. Validation with simulated datasets demonstrates that the proposed algorithm effectively reduces human intervention, avoids interference from outliers, and ensures more reliable and reasonable clustering outcomes. The results show that the proposed method not only maintains good consistency with traditional methods but also exhibits higher applicability, providing a reliable reference for dominant joint grouping in engineering applications.

Key words: density peak clustering, crow algorithm, validity evaluation indices, rock discontinuities, dominant grouping

中图分类号: