Page 97 - 《振动工程学报》2026年第2期
P. 97
第 39 卷第 2 期 振 动 工 程 学 报 Vol. 39 No. 2
2026 年 2 月 Journal of Vibration Engineering Feb. 2026
地 震 动 大 数 据 降 维 及 其 特 征 母 波 频 谱 分 析
王 晨, 俞瑞芳, 杨千里, 杨柳青
(中国地震局地球物理研究所,北京 100081)
摘要:通过数值拟合得到满足工程需求的设计地震动时程是弥补现有强震记录较少或在区域上分布不均匀的重要方法。由
于强震记录中包含着来自震源、传播路径及场地等信息,因此直接采用实际记录进行地震动拟合可以反映工程场地的地震地
质环境。现面临的问题是如何从大量的原始地震动记录中提取出数量合理且包含主要特征的地震动。本文考虑影响地震动
特性的主要因素,基于震级、距离和场地条件,对强震记录进行分组形成地震动数据集;引入主成分分析(principal component
analysis,PCA)算法,分析一个数据集的特征值、特征母波与原始地震动频谱特征之间的相关性;通过探讨影响特征母波频谱
特性的因素,提出了实现地震动数据降维的累计方差解释率(cumulative variance explanatory rate,CVE)阈值,并以土耳其地震
(Mw7.6)记录为目标进行了地震动拟合,进一步明确了地震动样本数量和 CVE 阈值的相关性,及其对地震动拟合结果的影
响,为实现大量地震动数据的降维提供了可靠的理论依据。
关键词: 主成分分析;数据降维;累积方差解释率;地震动拟合;土耳其地震
中图分类号:TU311.3 文献标志码:A DOI:10.16385/j.cnki.issn.1004-4523.202401013
Big data dimensionality reduction of ground motions and its spectral analysis
of characteristic waves
WANG Chen,YU Ruifang,YANG Qianli,YANG Liuqing
(Institute of Geophysics,China Earthquake Administration,Beijing 100081,China)
Abstract:Obtaining design ground motion time histories through numerical fitting to meet engineering requirements is an important approach
to compensate for the scarcity or uneven regional distribution of available strong motion records. Since strong motion records incorporate
information from the source, propagation path, and site conditions, using actual records directly for ground motion simulation can reflect the
seismological and geological environment of the engineering site. The current challenge lies in extracting a reasonable number of ground
motions that retain essential characteristics from a large set of original records. In this paper, the main factors influencing ground motion
characteristics are considered, and strong motion records are grouped into a dataset based on magnitude, distance, and site conditions. The
principal component analysis (PCA) method is introduced to analyze the correlation between eigenvalues, characteristic waveforms, and the
spectral characteristics of the original ground motions. By investigating the factors affecting the spectral properties of the characteristic
waveforms, a cumulative variance explained (CVE) threshold is proposed for dimensionality reduction of ground motion data. Ground motion
simulation is then conducted targeting records from the Turkey earthquake (Mw 7.6). The relationship between the number of ground motion
samples and the CVE threshold is further clarified, along with its influence on the simulation outcomes. This study provides a reliable
theoretical basis for the dimensionality reduction of large ground motion datasets.
Keywords:principal component analysis (PCA);data dimensionality reduction;cumulative variance explanatory rate (CVE);ground motion
simulation;Turkey Earthquake
工程实践中,用于结构抗震设计的地震动可采 拟合技术可以体现出地震发生的物理本质,定量表
用相似场地条件的实际地震动记录。然而,由于现 达复杂的断层破裂以及传播过程,实现从震源到场
有强震记录在区域上分布不均匀,往往很难获得满 地的全过程拟合。但将其应用于抗震设计还存在一
足结构设计需求的地震动时程,因此需要通过人工 些问题,比如地震学解析方法的优势是拟合远震和
拟合得到适用于不同地震地质环境的设计地震动。 具有较长周期的地震波,对于工程结构关注的地震
现有的地震动拟合方法中,以地震学方法为基础的 动主要频段的拟合效果并不好 [1-2] 。鉴于此,基于经
收稿日期:2024-01-08;修订日期:2024-04-09
基金项目:国家自然科学基金资助项目(51878627)

