弯剪型结构物理参数时域识别及地震动反演研究

发布时间：2019-05-05 15:21

【摘要】：本文以弯剪型结构为研究对象,针对输入未知、存在不同水平噪声和线性及非线性参数系统等情形下,研究结构物理参数时域识别及地震动反演问题。提出了求解非线性参数方程的SVDm LM法(Singular Value Decomposition-modified Levenberg-Marquardt),基于复合反演算法,进行非线性参数系统的参数识别和地震动反演;应用静力凝聚法,给出具有先验知识的转角重构公式;引入了矩形窗法,采用统计平均法,降低采样异常点对估计结果的影响。研究结果表明:本文提出的SVDm LM法能够有效的求解非线性参数方程;引入辅助条件,利用转角重构公式能够获得转角时程,且有效减弱了参数识别方程的病态程度,提高了参数识别和地震动反演精度;采用矩形窗法能够有效的减弱采样异常点的影响,从而降低估计误差。
[Abstract]:In this paper, considering the bending-shear structure as the research object, the time domain identification of physical parameters and the inversion of ground motion are studied under the conditions of unknown input, different horizontal noise, linear and nonlinear parameter systems and so on. The SVDm LM method (Singular Value Decomposition-modified Levenberg-Marquardt) is proposed to solve the nonlinear parameter equation. Based on the complex inversion algorithm, the parameter identification and ground motion inversion of the nonlinear parameter system are carried out. The formula of corner reconstruction with prior knowledge is given by using the static condensation method, and the rectangular window method is introduced to reduce the influence of sampling anomalies on the estimation results by using the statistical average method. The results show that the SVDm LM method proposed in this paper can effectively solve the nonlinear parametric equations. By introducing the auxiliary condition, the rotation time history can be obtained by using the angle reconstruction formula, and the ill-conditioned degree of the parameter identification equation can be effectively weakened, and the precision of parameter identification and ground motion inversion can be improved. The rectangular window method can effectively reduce the influence of sampling outliers and thus reduce the estimation error.
【作者单位】：中国地震局工程力学研究所;广州大学土木工程学院;广州大学工程抗震研究中心;
【基金】：国家自然科学基金(青年科学基金)项目(51208478)
【分类号】：P315.9;TU311.3

【参考文献】