Empirical mode decomposition (EMD) is proposed to identify linear structure under non-stationary excitation,and non-white noise coefficient is introduced under the assumption of random signals consisting of white nois...Empirical mode decomposition (EMD) is proposed to identify linear structure under non-stationary excitation,and non-white noise coefficient is introduced under the assumption of random signals consisting of white noise and non-white noise signals. The cross-correlation function of response signal is decomposed into mode functions and residue by EMD method. The identification technique of the modal parameters of single freedom degree is applied to each mode function to obtain natural frequencies, damping ratios and mode shapes. The results of identification of the five-degree freedom linear system demonstrate that the proposed method is effective in identifying the parameters of linear structures under non-stationary ambient excitation.展开更多
文摘基于协方差驱动随机子空间识别(covariance-driven stochastic subspace identification,SSI-COV)方法的模态参数识别具有强鲁棒性、高精度的优势,在结构工作模态分析中应用广泛。为保证模态参数识别的准确性,新近提出的基于随机子空间(stochastic subspace identification,SSI)的模态参数不确定性量化方法,可有效估计模态参数的方差,但由于其计算各中间变量时,需显式表示出Jacobian矩阵,导致矩阵运算维度高、计算效率低。为此,提出一种基于SSI-COV的模态参数不确定度高效计算方法。首先,计算振动响应相关函数的方差,通过奇异值分解(singular value decomposition,SVD),选取合适的奇异值截断阶数,由每阶奇异向量组装出多组Hankel矩阵的扰动。其次,根据一阶矩阵摄动理论,隐式计算SSI-COV算法各中间变量的一阶扰动,最终,由多组模态参数的扰动叠加计算出方差。最后,以桁架结构模型为例,采用所提方法辨识结构模态参数并计算模态参数方差,分析了Hankel矩阵维度及相关函数奇异值截断阶数对辨识结果的影响,结果表明计算得到的模态参数方差与蒙特卡洛仿真(Monte Carlo simulation,MCS)结果非常接近,且模态参数不确定度可作为剔除虚假模态的有效依据。
基金National Natural Science Foundation(No.19972016)for partly supporting this work
文摘Empirical mode decomposition (EMD) is proposed to identify linear structure under non-stationary excitation,and non-white noise coefficient is introduced under the assumption of random signals consisting of white noise and non-white noise signals. The cross-correlation function of response signal is decomposed into mode functions and residue by EMD method. The identification technique of the modal parameters of single freedom degree is applied to each mode function to obtain natural frequencies, damping ratios and mode shapes. The results of identification of the five-degree freedom linear system demonstrate that the proposed method is effective in identifying the parameters of linear structures under non-stationary ambient excitation.