利用CTCM优化K-means聚类研究微震震源稳健定位

Robust Location of Micro-Seismic Sources Based on CTCM-Optimized K-means Clustering

ES评分 0

DOI 10.14075/j.jgg.2025.12.408
刊名
Journal of Geodesy and Geodynamics
年,卷(期) 2026, 46(8)
作者
作者单位

1 成都锦城学院数学建模研究中心, 成都, 611731;
2 中国地震局地震研究所, 武汉, 430071;
3 武汉引力与固体潮国家野外科学观测研究站, 武汉, 430071

摘要
针对传统微震震源定位方法在噪声干扰下鲁棒性不足的问题,提出一种融合遗传算法(genetic algorithm, GA)全局搜索和部落竞争与成员合作算法(competition of tribes and cooperation of members algorithm,CTCM)优化K-means聚类的微震震源稳健定位框架。首先,基于到时差数学模型构建不同测点四四组合的最优化目标函数,通过GA算法遍历求解这些以未知震源三维坐标为自变量的三元二次函数,生成大量震源近似解;然后,引入CTCM对K-means聚类的初始聚类中心进行优化,将聚类结果的误差平方和(sum of squared error, SSE)作为最优化目标函数的因变量,以状态转移相关矩阵提升高维数据的聚类稳定性;最后,执行优化后的K-means得到多个聚类中心,以这些中心坐标的中位数和各点马氏距离构建抗差定位策略,并应用中位数绝对偏差(median absolute deviation, MAD)实现对异常点的自动剔除,其他聚类中心的加权算术平均值即为最终解。结果表明,该方法(GA-CTCM-K-means)在1组平面型仿真实验和2组真实微震案例下均能实现较好的异常剔除和定位效能,定位精度分别可达9.524 3 m、137.865 1 m、28.100 9 m,优于Geiger法、单纯形法、震源扫描法等传统方法,是一种具有一定应用价值的微震震源定位方案。
Abstract
To address the insufficient robustness of traditional micro-seismic source localization methods under noise interference, we propose a robust micro-seismic source localization framework that integrates genetic algorithm (GA) global search with CTCM-optimized K-means clustering. First, an optimization objective function is constructed based on the arrival time difference mathematical model for different four-by-four combinations of measurement points. The genetic algorithm is employed to traverse and solve these cubic quadratic functions with the unknown three-dimensional coordinates of seismic source as independent variables, generating many source approximate solutions. Second, the CTCM algorithm is introduced to optimize the initial K-means clustering centers. The sum of squared error (SSE) of clustering results is used as the dependent variable of optimization objective function, and the state transition correlation matrix is used to enhance the clustering stability of high-dimensional data. Finally, the optimized K-means yields multiple clustering centers. The robust localization strategy is constructed using the median of these center coordinates and the Mahalanobis distance of each point. The median absolute deviation (MAD) is applied to automatically remove outliers, and the weighted arithmetic mean of the remaining clustering centers is the final solution. The results show that GA-CTCM-K-means method achieves superior anomaly elimination and localization performance in one planar simulation case and two real micro-seismic cases, with localization accuracy reaching 9.524 3 m, 137.865 1 m, and 28.100 9 m, respectively, which is superior to traditional methods such as the Geiger method, simplex method, and source scanning method. This method represents a micro-seismic source localization scheme with practical application value.
关键词
微震震源定位; 到时差模型; CTCM算法; K-means聚类; 遗传算法; 稳健定位; 中位数绝对偏差
KeyWord
micro-seismic source localization; arrival time difference model; CTCM algorithm; K-means clustering; genetic algorithm; robust localization; median absolute deviation
基金项目
页码 1030-1038
  • 参考文献
  • 相关文献
  • 引用本文

陈国庆, 庞聪, 梁雨萱, 李尚东, 李查玮, 刘斯睿. 利用CTCM优化K-means聚类研究微震震源稳健定位 [J]. 大地测量与地球动力学. 2026; 46; (8). 1030 - 1038.

  • 文献评论

相关学者

相关机构