【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。更多还原

【Abstract】 For the shear-bending-type structure,the physical parameter identification in time domain and inversion of ground motion are studied in consideration of the white noise with different levels and linear / nonlinear parametric system without input. The SVD-m LM method( Singular Value Decomposition and modified Levenberg-Marquardt)is proposed for the nonlinear parametric system,and based on the hybrid inversion method,the parameter identification and ground motion inversion are performed. The formula of the reconstruction of rotation angle from the translation displacement is given by use of the Static condensation method. The rectangle window method and the statistical average algorithm are introduced for reducing the effect of the serious-noise-pollution samples on the estimation values. The results show that:( 1) The SVD-m LM method is effective for solving the nonlinear parametric equations;( 2) The reconstruction of rotation angle with the auxiliary conditions improves the precision of parameter identification and ground motion inversion;( 3) The application of the rectangle window method lowers the estimation errors.更多还原