基于XGBoost算法的工程场地实测和人工地震波时频特征分析与判别  

Time-frequency characteristic analyses of measured and artificial seismic waves using the XGBoost algorithm

作　　者：曹子龙 黄杜若 CAO Zilong;HUANG Duruo(State Key Laboratory of Hydroscience and Engineering,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学水沙科学与水利水电工程国家重点实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2022年第8期1330-1340,共11页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(52179134)。

摘　　要：近年来,人工智能方法快速发展,在许多工程问题中,逐渐引入具有良好预测能力和泛化能力的机器学习算法。该文考虑采用XGBoost人工智能方法,对工程场地实测和人工模拟地震波的时频规律特征进行深入探索,旨在解决地震波研究目前存在的资料缺乏与认识匮乏两大问题。采用的XGBoost算法优势在于人工智能方法的高性能计算能够完成传统计算方法难以实现的对大量数据的分析,从而发掘、重现地震波的时域和频域特征。在对实测和SIMQKE人工地震波的判别研究中,发现本算法对于二者的判别准确率能达到91%,进一步研究发现人工地震波与实测波差别主要体现在时频域特征的相关性上。该文有助于进一步认识地震波的时频特征,同时也对人工地震波模拟方法的发展具有重要意义。The rapid development of artificial intelligence and machine learning techniques that can predict and generalize data sets are gradually being introduced into many engineering projects.The XGBoost artificial intelligence method was used here to analyze the time-frequency characteristics of measured and artificially simulated seismic waves to resolve the two problems of the lack of data and the lack of understanding seismic waves.The advantage of the XGBoost method is that high-performance artificial intelligence methods can analyze large amounts of data that would be difficult by traditional calculational methods.The method can then be used to analyze the time-frequency domain characteristics of seismic waves.The algorithm accurately discriminated 91%of the measured and artificial SIMQKE seismic waves.Further research showed that the difference between the artificial seismic waves and the measured waves was mainly reflected in the correlation of the time-frequency domain characteristics.This study reveals the time-frequency characteristics of seismic waves that will facilitate the development of artificial seismic wave simulation methods.

关 键 词：人工地震波 XGBoost方法 智能监测 SIMQKE 小波包分析 

分 类 号：TU352.1[建筑科学—结构工程]