【作者】 孙晓丹；

【导师】 陶夏新；

【作者基本信息】 哈尔滨工业大学 ， 防灾减灾工程与防护工程， 2010， 博士

【摘要】 在大跨桥梁、长距离管线、大坝等长大空间建、构筑物的地震反应分析中,各输入点地震动之间的变化会产生重要影响,采用非一致的地震动输入,得到的结构内力比一致输入的有明显差别。需要考虑地震动的空间变化,采用多点激励,表达地震波到达各墩底的时间不同、各点所受到的激励之间并不完全相干、不同地点地震动振幅和频率成分的差别。统计结果表明,在几百米间距的尺度上,地表地震动的相关性与距离的大小和地震动的频率范围密切相关。在低于某一频率的范围内,相邻点地震动是完全相关的,可用确定性的波动理论来解释;在高于此频率的范围,相邻点的地震动是不完全相关的,显示出较高的随机性。从本质上讲,这个空间尺度上地震动相互间的相关性是由各点的地震动来源于同一个震源和十分相近的传播途径决定的。近场的地震动受震源的影响尤为显著,震源的埋深、尺寸、方位、运动类型和破裂过程,都强烈地影响近场地震动的特点和空间的分布。本文在我国近场地震动场估计最新进展的基础上继续推进,在四个环节上进一步改进,为发展一套大跨工程结构抗震分析的多点地震动输入方法,提供技术支撑。首先,指出了尽可能与地震活断层探查成果、区域地震学和地球物理学成果密切结合,确定震源模型中全局、局部参数的重要性。强调发震断层位置、产状和埋藏深度务必要由探查的结果确定;破裂长度、宽度和面积的确定,要充分重视区域观测数据的作用,即便不能直接观测到破裂面的参数,也应参考其他数据间接推测;凹凸体位置和尺寸的判断可以参考断裂分段资料。通过一工程实例,具体阐述了用截断正态分布将两类震源参数估计方法结合起来,同时表达不确定性的思路和技术方法。提出了利用试算点地震动反应谱与相应平均反应谱的残差,选择表达“平均”特征的震源模型的方法,以避免对地震动场空间特征表达的干扰。接着,对点源引起的高频地震动的随机合成方法进行了研究,并从震源幅值谱和相位谱两方面进行了改进:对震源幅值谱模型的改进体现在两方面,其一是建立了Masuda震源谱模型中频率项指数与破裂面积的关系,同时保持两个频率项指数之积为2.0,比较好地表达了拐角频率随破裂面扩展而逐步降低的趋势,克服了合成地震动对子源尺寸的依赖性;其二,为体现凹凸体是高频地震波的主要散射源,用子源地震矩与平均地震矩之比,进一步调整基于整个地震矩推算的拐角频率,同时将标度因子与错动量的不均匀分布联系起来。既保证了合成地震动幅值及远场辐射能量不受子源尺寸的影响,又对地震动估计的精度有所改善。最后借助北岭地震的实测数据,对两种改进效果进行了验证。对相位谱的改进体现在,强调一个点源引起地震动的相位谱应该与地壳速度结构、介质参数,尤其是场地与源点的几何关系密切相关,努力在随机合成中减少随机性。提出了将无限均匀介质中剪切位错源引起的位移所对应的加速度傅式谱的相位谱,与所估计的幅值谱结合,形成复谱,再通过傅立叶逆变换生成加速度时程的方法。算例检验的结果表明,得到的加速度时程的反应谱,能够表征基于大量随机相位谱合成的地震动的平均水平,亦借助北岭地震的实测数据,验证了改进方案的可靠性和有效性。本文改进,消除了随机相位谱对合成地震动场空间特征的干扰。随后,本文列出了地震学普遍坐标系下无限均匀介质中剪切位错源引起的位移解的详细计算过程,推导了普遍坐标系下点源方位、滑动矢量、法向量等计算公式。通过对比分析,研究了数值积分的简化及子源上升时间的计算问题,指出:当子源尺寸小于1km时,可考虑用子源面积直接乘上点位错源引起的位移解,简化数值积分;子源上升时间可根据震级—上升时间统计关系式确定,避免根据子源错动量估计。最后,本文研究了高、低频地震动在时域叠加中滤波和数值积分、微分的步骤,提出了一些参考原则:高频地震动位移和速度,可通过对高频加速度时程积分得到,不必通过修改地震动类型因子重复随机合成计算;依低频位移时程推算低频速度和加速度,低通滤波须在数值微分后进行,以防止出现振荡;最终的位移和速度时程可以通过直接叠加滤波后的高、低频位移和速度时程得到,也可以由叠加后的加速度时程积分得到。近场地震动相当复杂,强地震动场的估计需要持续不断地深入,本文的改进可以视为一步推进,在若干环节上有改善,但并不是彻底解决。结尾处,本文亦列出了若干有待进一步研究的问题。更多还原

【Abstract】 The variation of ground motion at support point is very important for the seismic analysis of structures with large dimension such as long-span bridge, long-distance pipeline, large dam, and so on. The internal force of the structure with non-uniform seismic input may be much different from that with uniform seismic input. The time for each support point when seismic wave arrives is different. The excitation at each support point is incompletely coherent. The amplitude and frequency content of ground motion at each support point is also different. Multi-support excitation considering the spatial variation of ground motion is needed to describe these above differences. According to the statistical results, the correlation of ground motion within a distance of several hundreds is closely related to the distance and the frequency content of ground motion. In the frequency range below a certain frequency, ground motions at adjacent points are completely coherent and can be represented by deterministic wave theory; for higher frequencies, ground motions at adjacent points are incompletely coherent and highly random. The essential reason of the correlation in this spatial scale is because ground motions at each point are from the same source and very close propagation path. Ground motion in the near field is significantly affected by the source. The depth, size, orientation, mechanism and rupture process of the source affect the characteristics and spatial distribution of near-fault ground motion field.Based on the latest progress of the estimation of near-field strong ground motion, this dissertation addresses four links of the method in order to provide better technical support of the multi-support excitation to the seismic analysis of long-span structures.The importance to determine the global and local parameters of the source model considering the results from active fault exploration, local seismological and geophysical research was pointed out. Three issues were emphasized: the location, attitude and depth of the causative fault must be determined according to the results from active fault exploration; the determination of rupture length, width, and area should consider the local observation data sufficiently, either the rupture plane parameters or some other parameters; the fault segmentation could provide a reference for the location and size of asperity. The procedure that combines two kinds of estimation by truncated normal distribution and represents uncertainties was demonstrated. In order to avoid distortion in the spatial pattern of the estimated ground motion field, choosing mean source model by evaluating the response spectral residuals at trial points was proposed.Two issues in stochastical approach for synthesizing the high-frequency ground motion from point source were studied. Three improvements on the two issues were made, two for source spectral model and one for phase spectrum.The first one of the two improvements on source spectral model was connecting the rupture area to the frequency exponents in Masuda source spectral model while keeping the product of the two frequency exponents as 2.0. This improvement describes the decrease of the corner frequency with rupture development and maintains the estimated ground motion independent of the fault-discretization scheme. In the second improvement, the corner frequency from the total seismic moment was adjusted by the ratio of the sub-source moment and the average moment to reflect that asperity was the major scattering source of high-frequency wave. The scaling factor was also adjusted to keep the estimated ground motion and far-field received energy independent of the fault-discretization scheme. With the second improvement, the precision of stochastical approach was enhanced significantly. The observed data of the Northridge earthquake was used to test the effects of the two improvements.The phase spectrum of ground motion from a point source should be related to crustal parameters, propagation medium parameters, and the geometrical relation between the source and the observation point. The uncertainties in the estimated ground motion from adopting random phase spectrum should be reduced. A new idea to improve phase spectrum was proposed. In the proposal, random phase spectrum was replaced by the phase spectrum of the acceleration Fourier spectrum from the displacement caused by shear dislocation source in infinite homogeneous space, the complex spectrum was formed by combining the new phase spectrum and the estimated amplitude spectrum, and the acceleration time history was obtained by the inverse Fourier transform of the complex spectrum. The new approach was tested with different fault mechanisms and rupture starting points. The test results showed that the response spectrum of the synthesized acceleration time history using the above proposal can represent the mean level of ground motions with different random phase spectra. The observed data of the Northridge earthquake was also used to test the effect of this improvement. This improvement eliminates the distortion of the random phase spectrum on the spatial pattern of estimated ground motion field.The detailed computation of the displacement caused by shear dislocation source in infinite homogeneous space was introduced. Equations for variables such as point source orientation, dislocation vector, and normal vector were deduced in seismological universal coordinate system. Two issues, the simplification of numerical integration and the computation of sub-source rise time, were discussed. The results show that: when the sub-source size is smaller than 1km, the numerical integration could be simplified by multiplying the sub-source area to the displacement from point dislocation source; the sub-source rise time should be determined by the relation between magnitude and rise time rather than by sub-source dislocation.The consequences of filtration, numerical differentiation, and numerical integration in the superposition of ground motions at low and high frequencies were studied. Some reference principles were proposed: the displacement and velocity in high frequencies can be obtained by the integration of acceleration instead of by changing the ground motion indicator; When computing the velocity and acceleration from low-frequency displacement, low-pass filtration must be performed after numerical differentiation to avoid the oscillation; The final displacement and velocity could be obtained either by the integration of the superposed acceleration or by the direct superposition of the filtered displacement and velocity.Ground motion in the near field is quite complicated. This dissertation only improved some links of the estimation of strong ground motion field but didn’t solve all problems. Issues that need to be researched further were proposed at the end of the dissertation.更多还原