【作者】 喻烟；

【导师】 李小军；

【作者基本信息】 中国地震局工程力学研究所 ， 结构工程， 2012， 博士

【摘要】 2008年5月12日，我国四川省汶川县遭遇了Mw7.9级大地震，强烈的地震动以及所带来的地质灾害，致使房屋、桥梁等结构物和构筑物倒塌不计其数，受灾地区人民生命财产和经济社会发展蒙受巨大损失。地质灾害和结构破坏的主要诱因是汶川地震区的强地面运动（又称强地震动）。地震动的研究是地震工程学中很重要的一部分。基于丰富的汶川地震动数据研究好汶川地震的地震动参数对我国的其他地震区、带的地震动研究有很大的借鉴意义。地震动是一个复杂的结合体，它是震源效应、地球介质衰减以及场地效应三者联合作用的产物。地震动的长周期成分具有明显的确定性，一般用表示定理给出。表示定理给出的地震动能较好地拟合地震动长周期波形。但是地球介质的复杂性导致根据物理模型求解的格林函数精度有限。同时，由于震源破裂过程的复杂性、地球介质的非均匀性、地形以及场地类别的多样性导致地震动的高频成分具有很强的随机性。因此，地震动场的预测和估计更多地依靠经验模型。经验模型主要分为两种，衰减关系模型和随机模型。对于强震动记录丰富的地区，衰减关系模型比较可靠，如美国加州、日本及中国台湾地区。我国强震动观测工作起步较晚，大震级的强震动数据较少，制约了我国在衰减关系模型的研究，导致大部分衰减关系并非根据我国强震动记录数据拟合而来，而为根据其他资料转换得到。然而，地震动有很强的地域性，这种转换方法的有效性需要未来大量地震数据的检验。对于强震动记录较少的地区，随机模型成为常用方法。国内应用随机模型的研究大部分参考国外的震源参数、衰减函数以及场地放大系数进行计算。但是，震源参数受目标区域地质构造、地壳应力场分布、发震断层类型等影响，衰减函数也受地壳厚度、地质构造等影响，由于地表土层参数对场地反应影响甚大，场地放大系数具有强烈的地域性特点。因此，采用其他地区的随机模型来模拟我国地震动场的方法需要用我国未来的强震动数据复核以及改进相应的模型参数。本论文分别从衰减关系和随机模型两方面拟合汶川地震强余震数据，建立了基于汶川地震强余震(Ms3.8~6.3)的地震动经验模型，并应用有限断层随机模型重演了汶川地震主震地震动场。本文采用的衰减关系模型为NGA Campbell-Bozorgnia模型，它用Vs30参数衡量场地特征。而汶川地震震感强烈的四川、甘肃地区大部分台站钻孔数据小于20m，为了获得较为准确的Vs30，本文基于147个四川、甘肃地区强震动观测台的建台报告，统计了我国西南地区浅硬土层场地的平均场地参数，得到了地区Vs30估计的经验模型。该模型可以为我国西南地区强震动记录数据结合国外的经验模型提供场地分类依据。结合Vs30的估计结果，利用NGA Campbell-Bozorgnia模型拟合汶川地震的强余震记录数据，得到了汶川地震强余震的地震动衰减关系。基于汶川余震序列的震源集中、台站数据密集等特点，分析了地震动衰减关系模型拟合的误差，将误差分解为震源误差、路径误差、场地误差及偶然误差。分析结果表明：基于单一路径的数据进行衰减关系研究，可以将误差的方差减小50%。根据13个汶川余震序列，反演了松潘-甘孜地块和四川盆地的Q模型、13个余震序列的震源参数以及浅硬土层场地和深厚土层的平均放大系数和高频衰减系数，构建了基于汶川地震强余震的点源随机模型。通过点源随机模型与衰减关系模型的对比，发现点源随机模型能更好估计近场地震动。论文最后应用有限断层模型重现了汶川地震主震地震动场。结果表明，有限断层模型能较好地模拟汶川地震主震近场地震动，可以为汶川地震中的结构破坏研究提供数据保障，也可用于预测将来该地区的大地震地震动场。更多还原

【Abstract】 On12May2008, an earthquake with Mw7.9occurred in Wenchuancounty, Sichuan province, China, which gave rise to many landslides,structural damage and collapses and resulted in widespread damage and agreat number of casualties. The main cause of those damage is strong groundmotion. The research of strong ground motions is a important part ofEarthquake Engineering. In particular, the research of strong ground motion inWenchuan earthquake is significant to other seismic areas of China.The ground motion is complex of source effect, crustal attenuation andsite condition effect.The long period component of ground motion isdeterministic and is given by representation theorem, which can simulate thelong period wave form very well. However, the precision of green fuction islimited by the heterogeneity and complexity of crust. Meanwhile, the highfrequency component of ground motion is stochastic, because of thecomplexity of source rupture process, the heterogeneity of crust and thevariety of site condition. As a result, empirical model is a more reliablemeasure for the prediction of ground motion.There are two empirical models for the prediction of ground motion,attenuation relationship model and stochastic model. Attenuation relationshipmodel is suitable for the region with sufficient strong motion records, such asEast America, Japan and Taiwan. Because chinese strong motion observationwork just initiate and the strong motion records is sparse, the attenuationrelationship for China are not derived from strong motion records, but fromother data, such as intensity data. Since the ground motion is characteristic ofregion, the reliability of this mothod need to be verified by the strong motionrecords in the future. For the region with limited strong motion data,stochastic model is a better choice. At present, many researcher usestochastic model, which are inverted from the strong motion records in WestAmerica or East Canada, to simulate the ground motion of Chinese earthquake.However, there are many aspects which result in ground motion differencesbetween the earthquake in China and other regions. Firstly, differentgeological structure and type of faulting classification will cause differentsource parameters. Secondly, the difference of crustal thickness andgeological structure will lead to different attenuation function. Thirdly, siteamplification factor is mainly determined by the surface soil, which isstrongly characrictic of region. As a result, the mothod, which use foreignparameter to simulate the ground moiton of earthquake in china, need Chinesestrong motion data to validate and update in the future.This thesis use attenuation relationship model and stochastic modelseparately to fit Wenchuan aftershocks (Ms3.8~6.3) and develop empirical ground motion model for Wnchuan aftershocks. At last, the ground moiton ofWenchuan mainshock are simulated by finite fault stochastic model.NGA Campbell-Bozorgnia model are used to fit Wenchuan aftershockdata, which take Vs30to describe the site classification. However, the drillingdepth of most strong motion stations in Sichuan and Gansu provinces is lessthan20meters. In order to extrapolate the profile of20meters to30meters,the empirical extrapolation model for the estimate of Vs30are derived from theconstruction report of147strong motion stations in Sichuan and Gansuprovinces. Based on the strong motion records of64wenchuan afterhshock,we get a new attenuation relationship, which consideres the site effect.Moreover, the estimate of Vs30can offer facility for the application of foreignmodel to Chinese data.Since the Wenchuan aftershocks are characteristic of concentratedSource location and multiple records at single station, the bias of attenuationrelationship regression can be decomposed into source bias, path bias,sitecondition bias and occasional bias. The results show that, if single path isconsidered, the sigma can be reduced to50%of the corresponding sigma forattenuation relationship regression.Based on the13wenchuan aftershocks, we invert the Q model separatelyfor the hanging wall and footwall of Longmenshan fault, stress dropes of the13afterhocks and site amplification factor for the shallow soil sites anddeep soil sites at Southwest China. Then the point source stochastic model isconstructed. Through the comparison of point source stochastic model andattenuation relationshiop model, it is noted that the point source stochasticmodel can simulate the near-fault ground motion better.Finally, finite fault stochastic model is used to simulate Wenchuanmainshock data. The fit of data show that the empirical model can simulatethe near-fault motion very well. The constructed model can offer artificialground motion for structure engineering and can be used to predict the groundmoion of large magnitude earthquake in Souchwest China.更多还原